Generation of Multiple Knowledge from Databases Based on Rough Sets Theory

Abstract

In this paper we present a new approach to generate multiple knowledge using rough sets theory. The idea is to generate several knowledge bases instead of one knowledge base for the classification of new object, hoping that the combination of answers of multiple knowledge bases result in better performance. Multiple knowledge bases can be formulated precisely and in an unified way within the framework of rough sets theory. Our approach is based on the reducts and decision matrix of the rough set theory. Our method first eliminates the superfluous attributes from the databases, next, the minimal decision rules are obtained through decision matrices. Then a set of reducts which include all the indispensable attributes to the learning task are computed, finally, the minimal decision rules are grouped to the corresponding reducts to form different knowledge bases. We attempt to make a theoretical model by using rough sets theory to explain the generation of multiple knowledge. The distinctive feature of our method over other methods of generating multiple knowledge is that in our method, each knowledge base is as accurate and complete as possible and at the same time as different from the other knowledge bases as possible. The test result shows the higher classification accuracy produced by multiple knowledge bases than that produced by single knowledge base.